[MeeGo-dev] Why "no upstream" is too expensive - was Re: RPM vs DEB, the FAQ item?

[David Greaves]

Auke Kok wrote:
> On 02/17/10 05:16, Jean-Christian de Rivaz wrote:
>> Learning that MeeGo have no upstream is for me a show stopper. It's
>> just a granted ticket for the hell.
OK thanks and goodbye.

Meanwhile lets see if we can discuss these misunderstandings or at least begin
to appreciate the various concerns and goals that different people have.

> this is a misconception. Just because (e.g.) moblin does not have "an
> upstream rpm repository like fedora" doesn't mean that there is no
> upstream at all. Upstream to moblin are the latest kernel sources, the
> tip of the Xorg tree, etc, and we follow these aggressively.

So I suggest that packaging a distro is essentially a QA process.

To me your statement seems to confuses package QA with distro QA.

Sure it's important to fix bugs - but much of the rest of a packaging exercise
is building against a consistent QA'ed set of 'frozen' libraries and utilities
and ensuring these various projects align. This means working with upstream to
find a good version/time to snapshot and to ensure that others in the ecosystem
will use the same base.

I don't think anyone doubts that creating a distro is a *LOT* of work - even
though most distros bend over backwards to minimse changes to upstream source
tarballs. So why are we not doing the open source thing and building upon the
efforts of Fedora/Debian/Suse etc ?

Any duplicated effort is wasted effort - and that's wasted money too.

So whilst I appreciate that we will have the flexibility to pursue our goals (as
yet unrealised) to be more efficient packagers - I'd say that improving the
efficiency of an inefficient process sounds .... sub-optimal.

Surely it would make more sense to work with an established distro-QA system and
improve the efficiency there?

Focus the auto-packaging on taking a consistent set of packages (from an
upstream distro) and MeeGo-ising the packaging - when there are problems go
upstream to the distro or go straight to the source. If we can repackage it
better then submit that to upstream.

A lot is made of Suse's involvement with Moblin and OBS and the Linux
Foundation.... well maybe we should look at openSuse a little closer? And if not
for MeeGo 1.0 then maybe for 1.1 or 1.2? Heck they are one of the best at i8n
(AFAIK).... they live in Germany which is barely a stone's throw from Helsinki.

David/lbt

-- 
"Don't worry, you'll be fine; I saw it work in a cartoon once..."